Introduction

Given all of the options for background subtraction in CalFUSE, deciding what to do in a given situation can be tricky.  Here's a guide for the perplexed.

Internal Logic of the Pipeline Routines

The background has two components, the detector dark count and the scattered light. The dark count is roughly constant across the detector but varies in intensity with time. The scattered light has considerable spatial structure, and this structure differs between day and night-time observations.  We have independent models for each of these components (day, night, and dark-count).  Properly scaling them is thus a problem with three unknown parameters.  We attempt to fit as many of these parameters as possible directly from the data.  Where that is impossible, we estimate one or more components and fit the remainder.

In its default mode, cf_make_ttag_bkgd uses the DNFLG array in the input photon list to divide the observation into day- and night-time portions and separately models the background for each.  From unilluminated regions of the detector, the program extracts both an empirical background spectrum and a single number representing the background intensity.  From the exposure time and the user-selected PHA threshold, it estimates the contribution to the background of the detector dark count and scales the appropriate scattered-light image to reproduce the observed background intensity.  To account for the variable dark count, the program repeats the sum, varying the assumed dark count until it achieves the best possible fit to the empirical background spectrum.  (Both the background sample regions and an initial estimate of the detector dark-count rate are stored in the parameter file bchr*.fit.) Once satisfied with the individual day and night-time background images, the program sums them to produce the final model background image.  The associated error image is (somewhat arbitrarily) set to 10% of the background (the errors are actually stored as variances), and the data-quality array is filled with 1s.

If the empirical background spectrum is very faint, cf_make_ttag_bkgd does not attempt to measure the detector dark count, but simply scales the mean dark-count rate by the the exposure time and fits a scattered-light model to the remaining background flux.  Users who require a more accurate background model may wish to combine data from multiple exposures using ttag_combine before running the pipeline (see below).

If the empirical background spectrum is very bright - as, for example, when nebular emission or a background star contaminates one of the other apertures (and thus the background-sample region) - no fit is performed.  Instead, the day and night components of the scattered-light model are scaled by the known day and night exposure times, the detector dark count is estimated from the mean count rate, and the three components are summed to produce a background image. This scheme is used for all histogram data files. Because histogram frames contain data only from the region about the desired extraction window, the scattered light cannot be measured from other regions of the detector.  Targets obtained in histogram mode are generally bright, so uncertainties in the background are less significant than for time-tagged data.

User-Controlled Options

We have seen cases in which nebular emission in other apertures contaminates the background-sample region but is not bright enough to trigger the bright-background switch in cf_make_ttag_bkgd.  To force the program to use this mode, set the keyword BKGDTYPE = -1 in the parameter file scrn*.fit.  Then, both the detector dark count and the scattered-light models will be scaled by the exposure time; no fit will be performed.

The background regions of short or night-only exposures may be too faint for CalFUSE to derive a full, two-component background model. In such cases, users may wish to combine data from multiple exposures into a single photon-list file using ttag_combine before running the pipeline. For best results, we recommend processing individual exposures through Step 6 (cf_add_shift) before combining the data. This way, corrections for Doppler shift and grating motion are derived using the orbital parameters most appropriate for each individual exposure.

For faint emission-line spectra, users may prefer not to subtract the background, but instead fit it as an independent spectral component.  To turn off background subtraction, set the keywords RUN_MKBK and RUN_BKGD to NO in the parameter file parm*.fit.  Both keywords must be changed.  Remember that there are two pipeline modules involved in the background: the first generates the background model, the second actually performs the subtraction.  When RUN_MKBK is set to NO,  cf_make_ttag_bkgd produces a background image of all zeros.  When RUN_BKGD is set to NO,  cf_bkgd copies input to output unchanged. Though not used by the background-subtraction module, the zero-value background image is needed by the both the distortion-correction and optimal-extraction routines.

Pulse-Height Screening

When reducing time-tag observations of faint sources for which the intrinsic background is a substantial fraction of the source flux, it is often advantageous to eliminate all events with low or high pulse heights, since these are primarily background events. (See The FUSE Observer's Guide for a complete discussion of the detector background and pulse-height distribution.) To do this, simply change the values of PHALOW and PHAHIGH in the appropriate scrn*.fit file. Reasonable values for Detector 1A are 4 and 16-20, respectively.

In practice, you should determine the optimal values of PHALOW and PHAHIGH for each observation. First, use fuse_scan to plot an image of the detector. Examine the pulse-height distribution for the region of the detector containing the spectral features of interest. (If you want the entire spectrum, be sure to exclude bright airglow lines.) Select values of PHALOW and PHAHIGH that span the broad peak of real photon events. For best results, extract spectra with PHALOW set to 3, 4, and 5 and see which value yields the highest signal-to-noise ratio.

The situation is complicated by the effects of gain sag on the detector. With time, the peak of the pulse-height distribution (for real photons) moves slowly to lower values. To compensate for this, we raise the detector voltage every six months or so, shifting the peak back to higher values. Data taken just before the voltage was raised will have a lower pulse-height peak than data obtained just after. That's why it's a good idea to examine the pulse-height distribution for each observation -- and each detector -- independently.

We are currently experimenting with schemes to improve the sensitivity of FUSE to faint sources by reducing sharply the range of accepted pulse heights. We find that the signal-to-noise ratio rises with PHALOW -- at least to a point, which seems to lie around PHALOW = 6-8 for LiF 1A. At this point, the detector background is significantly reduced, but at the cost of some fraction of the source photons, as well. It is thus likely that the standard FUSE flux calibration is inappropriate for data processed in this way. We will continue these experiments and provide more information as it becomes available.

Editing the Background Windows

It is possible to change the background sample regions used by cf_make_ttag_bkgd to derive the empirical background spectrum.  If you would like to explore this possibility, please send a note to fuse_support@pha.jhu.edu.

Time-Dependent Background Models

The scattered-light properties of the FUSE spectrograph appear to have evolved with time, reflecting changes in the detectors, the solar cycle, or both. To better model these changes, CalFUSE v2.1 employs a series of background-calibration files, each derived from blank-sky observations obtained during a particular time period ranging from six months to a year in length. For more information on the time dependence of the instrument background, please see Time-Dependent FUSE Calibration Effects, a CalFUSE White Paper.

Subtracting Your Own Background File

You may create a background file using whatever magic you like, but it must have the correct format or CalFUSE cannot use it. Luckily, the formats of all files used by CalFUSE are fully described in The CalFUSE Pipeline Reference Guide.

Inserting your background file into the pipeline is easy, but requires that you edit the shell script controlling the pipeline.  Simply add the following line after the call to cf_make_ttag_bkgd:

 Step 10     Make background file
 if !({$cfstat}) then
    cf_make_ttag_bkgd {$froot}_img.fit {$froot}_walk.fit \
                {$froot}_bkgd.fit       >>& $logfile
    set cfstat=$status
    cp my_model_bkgd.fit {$froot}_bkgd.fit 

 endif

Remember to copy, rather than move, your model background, especially if the pipeline is set to delete intermediate files!  (A symbolic link would work, too.)  If you comment out the call to cf_make_ttag_bkgd, your trailer file will be peppered with messages warning that the program has not been run.

Questions?

Please address questions to fuse_support@pha.jhu.edu.

straight to the top